Puzzled about Goal Seek

[Guest]

I have been teaching myself to use the FRP tool and so far things are going well; but a couple of questions:

1. What is the relationship between the annual retirement spending entry, and the goal seek function? When I increase spending, the goal seek, set at 99.9% - which displays as 100%, will often return a figure in detailed view that is much lower than the program returns at 100% success when goal seek is off. If I ask the question, "how much can I spend per year with a 100% success rate" given the same portfolio parameters, shouldn't the answer remain the same?

(I am entering current age and retirement age to be the same. Would that cause a problem?)

2. Can I adjust for 15 years during which I believe my budget may be $8000 less than entered, by entering $8000 as taxable savings for 15 years in the additional input screen? Or should I enter as no-cola income?

Thank you very much for your help, and for the program.

[jimr]

Hello,

Thanks for writing and for the good questions. I've tried to answer inline below..

What is the relationship between the annual retirement spending entry, and the goal seek function? When I increase spending, the goal seek, set at 99.9% - which displays as 100%, will often return a figure in detailed view that is much lower than the program returns at 100% success when goal seek is off. If I ask the question, "how much can I spend per year with a 100% success rate" given the same portfolio parameters, shouldn't the answer remain the same?

It sure would be better if the answers were the same

This seems to be a case of there being many roads to Dublin. With the target probability set to 100%, there's more than one "answer" to the goal-seek problem (eg more than one withdrawal rate that results in a probability of success of 100%). That's something I didn't code in. I probably should have kept you from entering anything greater than 99%.

I think if you use 95% or even 99%, it will work as expected.

But there is one caveat still. Projections made by Monte Carlo tools like FRP are fraught with imprecision. Thinking in terms of 100% probability of success isn't very realistic. Financial planning author William Bernstein once remarked that a Monte Carlo plan that results in an 80% probability of success and one with a 100% probability of success may in fact be equally likely to succeed in the real world. His point is that the estimation errors and other unknowns cause the output to have a guaranteed error of at least 20%.

I am entering current age and retirement age to be the same. Would that cause a problem?

That should be perfectly fine to do.

Can I adjust for 15 years during which I believe my budget may be $8000 less than entered, by entering $8000 as taxable savings for 15 years in the additional input screen? Or should I enter as no-cola income?

The answer to this depends on whether you really mean for the $8k amount to get reduced in real terms each year. If you use the income-no-cola option, the real amount will start at $8k, but will be reduced by inflation each year. That means that the actual spending reduction will be much less than $8k (real) by the end of the plan. If you just enter it as income, the amount of your expense reduction will keep up with inflation.

Usually the best way to check on things out is to take a good look at the year-by-year results in the detailed view tab. With that view you can really get a solid idea of what's going on year-to-year and overall. I find that view essential to making sure the planner is doing what I expect when it combines all the various options and overrides (as in additional inputs).

Please don't hesitate to post again if any of this doesn't make sense or if you have other questions.

Jim

[jimr]

One other caveat with goal-seek is that it's possible for the search to terminate without finding an answer.

Goal-seek uses the input retirement spending value (or rate of return) as a "seed" for the search. Starting with that seed, it runs simulations and adjusts the value up or down in an attempt to move the resulting probability of success in the right direction.

The adjustment made on each run of the simulation is a small fraction of the initial seed value. If the seed is too far away from the "right" answer, goal-seek may never get there. You can tell this is happening by checking if the resulting probability of success matches the target probability of success entered into the goal-seek settings.

fyi, Microsoft Excel has a goal-seek capability that works in much the same way.